Learning disabilities during childhood have adverse long-term consequences for academic and professional success and quality of life. Impaired learning can be particularly detrimental for children with autism spectrum disorder (ASD). Relative to their peers, children with ASD go on to achieve lower levels of post-secondary education, employment, and independent living. Yet, the cognitive, behavioral, and neural mechanisms underlying learning in children with ASD remain poorly understood. ASD is characterized by heterogeneous clinical presentations and cognitive abilities, likely resulting in highly variable learning profiles in affected children. Little is known about the neurobiology of learning in school-age children with ASD in any cognitive domain. In the current project period, we found heterogeneous patterns of math abilities in a large cohort of children with ASD, consistent with epidemiological reports indicating that 17-40% of children with ASD display lower-than-expected math achievement scores. We also found novel evidence that math abilities are associated with restricted and repetitive interests and behaviors (RRIB), a core clinical symptom of ASD. We propose, in this renewal, to leverage our innovative and high-impact line of research to investigate heterogeneity in learning and brain plasticity, and its links to RRIB and cognitive inflexibility, in children with ASD. Using a theoretically motivated cognitive training protocol, state-of-the-art brain imaging, and advanced multivariate computational techniques, we propose to test the hypotheses that (i) children with ASD and low math abilities (LMA-ASD) will show different learning profiles relative to children with ASD and high math abilities (HMA-ASD) and typically developing (TD) children and (ii) compared to the HMA-ASD and TD groups, the LMA-ASD group will demonstrate weaker plasticity in two distinct brain systems important for math learning: the visuospatial number system, anchored in the intra-parietal sulcus and fusiform gyrus, and the declarative memory system, anchored in the medial temporal lobe. We will also investigate whether RRIB and cognitive inflexibility have a negative influence on learning in ASD, and determine the extent to which these effects are mediated by aberrant functioning of the salience network, a prefrontal cognitive control system anchored in the anterior insula and anterior cingulate cortex. The proposed work will provide important new insights into the neurocognitive basis of heterogeneous learning profiles in children with ASD and is highly relevant to the mission of the NIH ?Research on Autism Spectrum Disorders? (PA-16-388). Identifying behavioral and neural sources of heterogeneity in learning and their links to clinical symptoms in a quantitatively rigorous manner will have significant implications for informing the etiology of ASD and more critically, for optimizing learning in affected children.